Those current requirements are quite easy, and your software can make a great saving too if well written. If your design requirements allow it, then the MCU can power down for most of the time, just powering up from an internal timer for the sensor read and transmit. Also, you might be able to turn off the sensor power supply if they have a quick start-up time, and you rewire the power supply so that the 3.3V regulator runs from the battery as well. Anyway...
What is your transmitting:not-transmitting ratio? That will give you the average current. Let's say for argument that it's 25%. Let's also say that for 50% of the time the MCU will be put into power down mode by your software, waiting to be awakened by a software timer for the next data burst. You might have other things that need some current too though, for example pull-up resistors etc. I'll leave those to you.
So, 50% at 1.6µA, 25% at 20mA and 25% at 80mA. On average then we have a current consumption of 0.5*0.0016 + 0.25*20 + 0.25*80 = 25mA. That's just for the MCU. We still need a regulator capable of 80mA (at least) peak current, but for your power dissipation and battery life calculations the average is only 25mA.
For the sensors, without more information on possible power-down then we need to look at the worst case. We'll call that a total of 10mA on the 5V line.
Wired the way you have it then, the 5V regulator, which also passes the current for the 3.3V regulator, will have 90mA peak and 35mA average going through it. The 3.3V regulator will just have its own 80mA peak and 25mA average. You have a very wide choice of regulators with current requirements being quite low.
I would still very much advise a low-dropout regulator for the 5V, and you need one also (as you already have in the diagram) for the 3.3V.
Battery life: Alkaline cells have a greater capacity than NiMH, but their voltage drops off continually throughout use so they often 'fail' before NiMH in a circuit that needs a certain voltage. NiMH keeps almost a constant 1.2V until almost exhausted. Given 2000mAh for an average AA NiMH cell, then with the averages that we have calculated above, an NiMH battery will last for 2000/35 = 57 hours.
Not long!
If you transmit requirements allow it though, you could make far greater savings on current by running the MCU at power down for most of the time, like 90%, and just waking up periodically to do a sensor-read and transmit.
You could even look at using a 5V regulator with an on/off control input, or using a MOSFET switch to turn the 5V rail on and off. Then, if you connect the 3.3V regulator directly to the battery so that the MCU always has power, you can turn off the 5V rail to the sensors to save their power during power-down mode. That depends on their start-up time as well though.
also, a switching regulator as suggested would be a good improvement giving perhaps 50% extra battery life (with 7.2V NiMH stack) at a quick guess; more if you use a higher voltage battery.
I hope some of that helps...